CN113792758B - Rolling bearing fault diagnosis method based on self-supervision learning and clustering - Google Patents
Rolling bearing fault diagnosis method based on self-supervision learning and clustering Download PDFInfo
- Publication number
- CN113792758B CN113792758B CN202110949934.XA CN202110949934A CN113792758B CN 113792758 B CN113792758 B CN 113792758B CN 202110949934 A CN202110949934 A CN 202110949934A CN 113792758 B CN113792758 B CN 113792758B
- Authority
- CN
- China
- Prior art keywords
- data
- domain
- self
- rolling bearing
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005096 rolling process Methods 0.000 title claims abstract description 42
- 238000003745 diagnosis Methods 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 15
- 230000009466 transformation Effects 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 31
- 238000013508 migration Methods 0.000 claims description 18
- 230000005012 migration Effects 0.000 claims description 18
- 238000013507 mapping Methods 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 11
- 238000013526 transfer learning Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010892 electric spark Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a rolling bearing fault diagnosis method based on self-supervision learning and clustering, which comprises the steps of firstly, based on the recognition of different time-frequency transformation types of bearing data by a self-supervision learning network, extracting the bottom non-deflection characteristics of two-domain data, then training source domain data in a rolling bearing fault diagnosis data set in a supervision learning mode, and predicting by using a source domain supervision learning network to obtain an initial pseudo tag of target domain data in the rolling bearing fault diagnosis data set; secondly, generating a pseudo tag and a probability value thereof based on network prediction, clustering the characteristics of the target domain data extracted from the self-supervision network by using a K-means algorithm in consideration of the self-distribution characteristics of the target domain data, and updating the pseudo tag and the probability value according to a strong cluster rule; and finally, setting the updated probability value as the confidence coefficient of the corresponding sample pseudo tag, wherein the overall average value is used as the overall confidence coefficient of the class, so that the availability of the pseudo tag is further improved, and the self-adaptive fault diagnosis in the unsupervised field is realized.
Description
Technical Field
The invention belongs to the field of fault diagnosis, and particularly relates to a rolling bearing fault diagnosis method.
Background
Along with the development of industrialization, an intelligent fault diagnosis system is also more and more critical, and a rolling bearing is a key component of a transmission device of many rotary mechanical equipment, and is one of the most parts with faults because of the complex and various operating environments, and in the mechanical operation process, phenomena such as overload, fatigue, abrasion, corrosion and the like are likely to cause bearing damage. Once the fault occurs, the normal operation of the equipment is affected by the light fault, serious safety accidents are caused by the heavy fault, and huge economic loss and casualties are caused, so that the running state of the rolling bearing is very meaningful to monitor and diagnose in order to improve the safety of the rolling bearing and prevent the occurrence of the accidents.
Current fault diagnosis methods for rolling bearings are mainly divided into two categories: model-based and data-based. Random factors and noise of a model-based fault diagnosis system in an actual equipment working environment are difficult to estimate in advance, so that an accurate and effective mathematical model is difficult to construct. In recent years, the fault diagnosis method based on data driving is supported by sufficient data basis and theory, and is mainly divided into two development routes, namely the traditional time-frequency analysis based on collecting vibration signals and machine learning algorithm for realizing fault identification, and the current end-to-end algorithm based on a deep neural network. The traditional fault diagnosis algorithm is limited in feature extraction capability for large data with large data quantity, higher data dimension and stronger nonlinear relation, and cannot completely extract essential features of the data. However, the good performance of deep neural networks in fault diagnosis requires two key preconditions: rich tagged data, and independent co-distribution between training and test data. However, in most industrial scenes, it is difficult to obtain enough tag data, and in order to solve this problem, migration learning is proposed, and as one of the representative methods of migration learning, domain adaptation is to mine domain non-offset features, bridge distribution differences between source domains and target domains, so as to migrate knowledge from a labeled source domain to an unlabeled target domain, and assist the target domain in completing classification tasks.
The occurrence of transfer learning solves the core problem that the deep neural network cannot be applied to large-scale practical application, and accelerates the floor implementation of the intelligent fault diagnosis system. However, in the current fault diagnosis algorithm, when the difference of the source domain and the target domain data distribution in the fault diagnosis of the rolling bearing is larger, the network which depends on the source domain data training tends to be more biased towards the source domain, so that the target domain features extracted by the network have certain deviation, and the feature distribution of the rolling bearing fault data which is adaptive to two domains in the state is difficult to extract better self-adaptive features. And the precision of condition distribution depends on the pseudo tag, and the reliability of the pseudo tag cannot be ensured by a pre-training network trained based on source domain data, so that the fault diagnosis precision of the rolling bearing is not high.
Disclosure of Invention
In order to solve the technical problems mentioned in the background art, the invention provides a rolling bearing fault diagnosis method based on self-supervision learning and clustering.
In order to achieve the technical purpose, the technical scheme of the invention is as follows:
a rolling bearing fault diagnosis method based on self-supervision learning and clustering comprises the following steps:
(1) Dividing the collected vibration signals of the driving end and the fan end of the rolling bearing into source domain data and target domain data, performing time-frequency conversion on the label-free data of the two domains, and obtaining a rolling bearing fault diagnosis data set D for self-supervision learning self Randomly initializing model parameters of a self-supervision learning network and based on a data set D self Training a self-supervised learning network, repeating the minimization of the loss function to update the model parameters θ of the self-supervised learning network 1 Until the loss function converges;
(2) Migrating model parameters θ of a partially self-supervised learning network in a locked form 1 To the source domain supervised learning network, the data set D self The source domain data and the source domain label in the network are used as the input of the source domain supervision and learning network to obtain the initialization parameters thereof, and the minimized loss function is repeated to update the model parameters theta of the source domain supervision and learning network 2 Until the loss function converges;
(3) Initial pseudo tag for predicting target domain data in rolling bearing fault diagnosis data set through source domain supervised learning networkCorresponding confidence level->Wherein n is t Representing the number of unlabeled data in the target domain, < >>The label corresponding to the j-th data in the target domain data is represented, and the model parameter theta of the partial source domain supervised learning network is migrated in a locked mode 2 To a two-domain transfer learning network, the data set D self The source domain data, the source domain label, the target domain data and the target domain initial pseudo label in the two-domain migration network are used as the input of the two-domain migration network to obtain the initialization parameter theta 3 ；
(4) Computing data set D self Collecting the edge distribution distance and the condition distribution distance of the target domain and the source domain and the self-adaptive weight factor of the condition distribution, minimizing the loss function to update the model parameter theta of the two-domain migration network 3 ；
(5) Updating data set D through optimized two-domain transfer learning network self Pseudo tag for medium target domainExtracting domain non-offset characteristics of target domain data, clustering the extracted domain non-offset characteristics based on a K-Means algorithm, and generating corresponding by using a clustering result and a pseudo tagDictionary mapping, generating corresponding strong clusters based on the dictionary mapping and the strong cluster rule, and further updating pseudo tags in the rolling bearing data set by using the strong clusters>Generating a corresponding data set D according to the prediction probability of the pseudo tag and the clustering result self Confidence of middle target Domain data class +.>
(6) Repeating steps (4) and (5) until the loss function converges or the data set D self The pseudo tag of the target domain is not updated any more; and inputting the target domain test data into the updated two-domain migration network to obtain the accuracy rate of the rolling bearing fault identification.
Further, the expression of the loss function in step (1) is as follows:
wherein loss (x i θ) is a loss function, K is the number of time-frequency conversion methods adopted, and g (|y) represents rolling bearing fault diagnosis original data { x } i Some kind of time-frequency transformation corresponding to F 1 Y (g(x i |Y)|θ 1 ) Is to output data through self-supervision learning network F 1 Generating probability values for predictive categories, θ 1 Representative self-supervised learning network F 1 Is a parameter of (a).
Further, the expression of the loss function in step (2) is as follows:
wherein,as a loss function, n s For the number of data in the source domain>For the tag corresponding to the i-th data in the source domain data,>source field data in the fault diagnosis data set for a rolling bearing>Learning network F through source domain supervision 2 Generating probability values for predictive categories, θ 2 Is a source domain supervised learning network F 2 Is a parameter of (a).
Further, in step (3), the confidence levelThe expression of (2) is as follows:
wherein,is the target domain data in the rolling bearing fault diagnosis data set +.>All classes C T The total number of samples of a certain class k, +.>For the probability that the sample belongs to the corresponding class, +.>The expression of (2) is as follows:
wherein,target field data in the fault diagnosis data set for the rolling bearing>Migrating a network F over two domains 3 Generating probability values θ for predictive categories 3 Is a two-domain migration network F 3 Is selected as the pseudo tag
Further, in step (4), an edge distribution distance D between the source domain and the target domain in the rolling bearing failure diagnosis data set is calculated M And conditional distribution distance D C ：
Wherein H is k The spatial mapping of the finger-hilbert is performed,is the source domain data in the rolling bearing fault diagnosis data set>All classes C T The total number of samples, sigma, of a certain class k (k) The confidence coefficient of the kth sample is that the initial value is obtained in the step (3)Subsequent sigma (k) Confidence after updating the strong cluster rule;
calculating an adaptive weight factor mu of the condition distribution:
wherein,for D M Is an unbiased estimate of->For D C Is an unbiased estimate of (1);
calculating a loss function:
wherein L is total (x；θ 3 ) As a loss function, lambda is a regularization parameter,adapting distance for joint distribution of two-domain data:
further, the strong cluster rule is that the dictionary mapping is generated by using the network prediction pseudo tag and the pseudo tag generated by clustering in the same batch, a threshold value alpha and a minimum sample number n=batch/n·β of the dictionary mapping are set first, wherein batch is the batch size, N is the class number, β is a parameter for determining the minimum sample size, if the ratio of the ith class of the pseudo tag generated by clustering to the jth class of the pseudo tag generated by network prediction is greater than the threshold value alpha and the sample number of the dictionary mapping is greater than the minimum sample number N, the pseudo tag i generated by clustering and the confidence are regarded as the pseudo tag and the confidence of the network, and the confidence is in the step (5)Otherwise pseudo-label of network predictionThe signature j and confidence are used as pseudo-labels and confidence of the network.
Further, in step (5), the confidence levelThe expression of (2) is as follows:
wherein,is the number of samples predicted to be other tags than the cluster tag.
The beneficial effects brought by adopting the technical scheme are that:
the invention extracts the bottom layer 'unbiased' characteristic of the two-domain data more fairly and purely in an unsupervised form through the identification of different time-frequency transformation types of the bearing data based on the self-supervised learning network. The invention utilizes a clustering algorithm to cluster the data characteristics extracted from the self-supervision network, updates the pseudo tag and the confidence coefficient thereof according to the set 'strong cluster' rule, and improves the accuracy of the pseudo tag.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a schematic diagram of self-supervised learning dataset preprocessing in the present invention;
FIG. 3 is a schematic diagram of pseudo tags generated by network prediction and clustering in the present invention.
Detailed Description
The technical scheme of the present invention will be described in detail below with reference to the accompanying drawings.
The invention designs a rolling bearing fault diagnosis method based on self-supervision learning and clustering, which comprises three links as shown in figure 1: 1) Collecting data; 2) Training a model; 3) And (5) fault diagnosis.
And (3) data acquisition: different data platforms are collected, and one data platform easy to collect is set as a source domain to collect data with labels. Setting another poor-condition data platform as a target domain, and collecting data without labels; this example collected a sensor dataset from U.S. kesixi Chu Da (CWRU) and a sensor dataset from the university of wortmann in canada (BV), wherein the induction motor bearing model of the CWRU dataset was SKF-6205-2RS. The vibration signals of the bearing are collected by an acceleration sensor, and are respectively arranged at the driving end and the fan end of the motor shell, and the sampling frequency of the signals is 12kHz. The deployment of BV data set equipment is similar to CWRU data set equipment, and the main difference is that the test bearing model is MFS-PK5M, and the sampling frequency of the signal is 20kHz. In order to simulate the bearing faults, single-point faults are introduced on the inner ring, the rolling bodies and the outer ring of the bearing by means of electric spark machining. In the embodiment, a part of data is selected from the CWRU data set and the BV data set to construct the data set of the experiment, the practical manifestations of health, inner ring damage and outer ring damage under different rotating speeds are tested through the experiment, and corresponding vibration and acceleration signals are collected. In order to further simulate a real scene, the characteristics of the CWRU data set are combined, and data of different loads of the platform are mixed with data of different fault degrees to form a data set G-J, wherein the specific details of the data set G-J are shown in Table 1. In order to test the migration performance of the migration learning algorithm across platforms, migration tasks are set to be G-I, G-J, H-I, H-J, I-G, I-H, J-G, J-H, wherein 400 groups of 500 samples in each group are taken as training sets, the rest are taken as test sets, G-I is taken as an example here to illustrate, a data set A (mechanical vibration data collected under 1 HP) is taken as a source domain, a data set G is taken as a target domain, at the moment, a data set I is provided with a label, and the data set G is not provided with the label. In the training process, training data of data sets G and I are selected, and a corresponding test set is selected for testing during testing, so that the test data is ensured not to participate in training.
TABLE 1G-I settings for multi-platform dataset
Model training: the method comprises three parts, namely, first, self-supervision network learning: as shown in fig. 2Performing multiple time-frequency preprocessing on the two-domain unlabeled data to obtain a data set D for self-supervision learning self Randomly initializing model parameters of the self-supervision learning network and training the self-supervision network D based on the data set self Repeatedly minimizing a loss function to update model parameters θ of a self-supervised learning network 1 Until the loss function converges; second, source domain supervised learning: migrating model parameters θ of a partially self-supervising network in the form of lockers (frozens) 1 To the source domain supervised learning network to obtain the initialization parameter theta 2 Updating a minimization loss function to update model parameters θ of a source domain supervised learning network 2 Until the loss function converges; third,: training of two-domain migration networks: initial pseudo tag for predicting target domain data through source domain supervised learning networkCorresponding confidence levelMigrating model parameters θ of a partially self-supervising network in the form of lockers (frozens) 2 To a two-domain transfer learning network to obtain the initialization parameter theta 3 Calculate the two-domain distribution distance +.>And->And a conditional distribution adaptive weighting factor μ, minimizing a loss function to update model parameters θ of the target network 3 . As shown in FIG. 3, the pseudo tag is updated by the optimized two-domain transfer learning network>Extracting domain non-offset characteristics of target domain data, clustering the extracted target domain characteristics based on a K-Means algorithm, generating corresponding dictionary mapping by utilizing a clustering result and a pseudo tag, and generating corresponding ' based on the dictionary mapping and a ' strong cluster ' rule "Strong clusters). And further update the pseudo tag with it +.>Generating confidence level of corresponding category according to the prediction probability of pseudo tag and clustering result>Updating minimizes the loss function until the loss function converges or the pseudo tag is no longer updated. Fault diagnosis: and testing the training to a convergence model by adopting target domain data to further verify the performance of the algorithm.
Table 2 shows the comparison between the method (DASSL-FC) proposed in this embodiment and other deep migration learning algorithms, although cross-platform acquisition results in a larger difference between two-domain data distribution, the algorithm proposed in the present invention reasonably uses the characteristics of the two-domain data itself to perform feature extraction through self-supervision learning, and by combining with adapting the joint probability distribution of the two domains at the same time, the domain non-offset feature can be extracted more effectively, and the confidence coefficient is improved by using the clustering and the "strong clustering" rule, so that the optimal prediction performance is obtained, and the average prediction accuracy of each migration experiment can reach about 85%. While CNNs have large differences in data distribution in the face of two domains, they cannot achieve good migration performance. The method DAN and DDC which are only suitable for edge distribution are not good, and JAN is very excellent in partial migration tasks, but the precision of processing other tasks is not high, and obviously the generalization of the algorithm is not good enough. theDAFDM,DAFDM-aandDAFDM-acalgorithmsignorethetargetdomain,resultinginatrainednetworkthatismorebiasedtowardsthesourcedomain,andthealgorithmislessgeneralized.
Table 2 Classification precision of Cross-platform migration tasks
The embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited by the embodiments, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the present invention.
Claims (4)
1. The rolling bearing fault diagnosis method based on self-supervision learning and clustering is characterized by comprising the following steps of:
(1) Dividing the collected vibration signals of the driving end and the fan end of the rolling bearing into source domain data and target domain data, wherein the source domain data comprises labeled source domain data and unlabeled source domain data, the target domain data is unlabeled target domain data, performing time-frequency transformation on the unlabeled data of the two domains, and obtaining a rolling bearing fault diagnosis data set D of self-supervision learning self Randomly initializing model parameters of a self-supervision learning network and based on a data set D self Training a self-supervised learning network, repeating the minimization of the loss function to update the model parameters θ of the self-supervised learning network 1 Until the loss function converges;
(2) Migrating model parameters θ of a partially self-supervised learning network in a locked form 1 To the source domain supervised learning network, using the source domain data and the source domain label as input of the source domain supervised learning network to obtain initialization parameters thereof, and repeatedly minimizing a loss function to update model parameters theta of the source domain supervised learning network 2 Until the loss function converges;
(3) Initial pseudo tag for predicting target domain data in rolling bearing fault diagnosis data set through source domain supervised learning networkCorresponding confidence level->Wherein n is t Representing the number of unlabeled data in the target domain, < >>Initial pseudo tag corresponding to jth data in target domain data is represented, and model parameter theta of partial source domain supervised learning network is migrated in a locked mode 2 To a two-domain transfer learning network, taking the two-domain labeled data set as the input of the two-domain transfer learning network to obtain the initialization parameter theta 3 The method comprises the steps of carrying out a first treatment on the surface of the The two-domain tagged data set comprises: source domain data, source domain labels, target domain data and target domain initial pseudo labels;
(4) Calculating the edge distribution distance and the condition distribution distance of the target domain and the source domain in the two-domain labeled data set and the self-adaptive weight factor of the condition distribution, and minimizing the loss function to update the model parameter theta of the two-domain migration network 3 ；
(5) Updating pseudo labels of target domains in two-domain labeled dataset through optimized two-domain transfer learning networkExtracting domain non-offset characteristics of target domain data, clustering the extracted domain non-offset characteristics based on a K-Means algorithm, generating corresponding dictionary mapping by using a clustering result and a pseudo tag, generating corresponding strong clusters based on the dictionary mapping and a strong cluster rule, and further updating the pseudo tag ++in the rolling bearing data set by using the strong clusters>Generating confidence level of target domain data category in the data set corresponding to the two domains with labels according to the prediction probability of the pseudo labels and the clustering result>
(6) Repeating the steps (4) and (5) until the loss function converges or the pseudo tag of the target domain in the two-domain tagged dataset is not updated; inputting the target domain test data into the updated two-domain migration network to obtain the accuracy rate of the rolling bearing fault identification;
in step (3), confidence levelThe expression of (2) is as follows:
wherein,is the target domain data in the rolling bearing fault diagnosis data set +.>All classes C T The total number of samples of a certain class k, +.>For the probability that the sample belongs to the corresponding class, +.>The expression of (2) is as follows:
wherein,target field data in the fault diagnosis data set for the rolling bearing>Learning network F through source domain supervision 2 Generating a probability value of the predicted category, wherein the category with the highest probability is selected as a pseudo tag +.>
The strong cluster rule is that dictionary mapping is generated by using the network prediction pseudo tags of the same batch and the pseudo tags generated by clustering, firstly, a threshold value alpha and a minimum sample number n=batch/N.beta of the dictionary mapping are set, wherein batch is the batch size, N is the class number, beta is a parameter for determining the minimum sample size, and if the dictionary mapping proportion of the ith class of the network prediction pseudo tags generated by clustering and the jth class of the network prediction pseudo tags is larger than the threshold value alpha and the sample number of the dictionary mapping is larger than the minimum sample number N, the pseudo tags i generated by clustering and the confidence are used as the pseudo tags and the confidence of the network, and the confidence is in the step (5)Otherwise, taking the pseudo tag j and the confidence coefficient of the network prediction as the pseudo tag and the confidence coefficient of the network;
in step (5), confidence levelThe expression of (2) is as follows:
wherein,is the number of samples predicted to be other tags than the cluster tag.
2. The self-supervised learning and clustering-based rolling bearing fault diagnosis method according to claim 1, wherein the expression of the loss function in step (1) is as follows:
wherein loss (x i θ) is a loss function, K is the number of time-frequency transform methods employed,representing rolling bearing fault diagnosis raw data { x } i Some kind of time-frequency transformation corresponding to F 1 Y (g(x i |Y)θ 1 ) Is to output data through self-supervision learning network F 1 Generating probability values for predictive categories, θ 1 Representative self-supervised learning network F 1 Is a parameter of (a).
3. The self-supervised learning and clustering-based rolling bearing fault diagnosis method according to claim 1, wherein the expression of the loss function in step (2) is as follows:
wherein,as a loss function, n s For the number of data in the source domain>For the tag corresponding to the i-th data in the source domain data,>source field data in the fault diagnosis data set for a rolling bearing>Learning network F through source domain supervision 2 Generating probability values for predictive categories, θ 2 Is a source domain supervised learning network F 2 Is a parameter of (a).
4. The method for diagnosing a rolling bearing failure based on self-supervised learning and clustering as recited in claim 1, wherein in step (4), a two-domain band is calculatedEdge distribution distance D of source domain and target domain in tag dataset M And conditional distribution distance D C ：
Wherein,finger Hilbert spatial mapping, +.>Is the source domain data in the rolling bearing fault diagnosis data set>All classes C T The total number of samples, sigma, of a certain class k (k) For the confidence level of the kth sample, the initial value is +.>Subsequent sigma (k) Confidence after updating for strong cluster rule, +.>Target field data in the data set with labels for two fields +.>Migrating a network F over two domains 3 Generating a probability value for the predicted category;
calculating an adaptive weight factor mu of the condition distribution:
wherein,for D M Is an unbiased estimate of->For D C Is an unbiased estimate of (1);
calculating a loss function:
wherein,as a loss function, λ is a regularization parameter, +.>Adapting distance for joint distribution of two-domain data:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110949934.XA CN113792758B (en) | 2021-08-18 | 2021-08-18 | Rolling bearing fault diagnosis method based on self-supervision learning and clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110949934.XA CN113792758B (en) | 2021-08-18 | 2021-08-18 | Rolling bearing fault diagnosis method based on self-supervision learning and clustering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113792758A CN113792758A (en) | 2021-12-14 |
| CN113792758B true CN113792758B (en) | 2023-11-07 |
Family
ID=78876105
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110949934.XA Active CN113792758B (en) | 2021-08-18 | 2021-08-18 | Rolling bearing fault diagnosis method based on self-supervision learning and clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113792758B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114331214A (en) * | 2022-01-21 | 2022-04-12 | 山东大学 | Domain-adaptive bearing voiceprint fault diagnosis method and system based on reinforcement learning |
| CN114972920B (en) * | 2022-05-30 | 2024-03-12 | 西北工业大学 | Multi-level non-supervision field self-adaptive target detection and identification method |
| CN116150635B (en) * | 2023-04-18 | 2023-07-25 | 中国海洋大学 | Rolling bearing unknown fault detection method based on cross-domain relevance representation |
| CN116383737B (en) * | 2023-06-05 | 2023-08-11 | 四川大学 | Rotary machine fault diagnosis method and system based on cluster comparison learning |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102039138B1 (en) * | 2019-04-02 | 2019-10-31 | 주식회사 루닛 | Method for domain adaptation based on adversarial learning and apparatus thereof |
| CN110705525A (en) * | 2019-10-24 | 2020-01-17 | 北京建筑大学 | Method and device for diagnosing rolling bearing fault |
| CN111144565A (en) * | 2019-12-27 | 2020-05-12 | 中国人民解放军军事科学院国防科技创新研究院 | Self-supervision field self-adaptive deep learning method based on consistency training |
| CN111144458A (en) * | 2019-12-13 | 2020-05-12 | 重庆交通大学 | Method for identifying mechanical faults under different working conditions of subspace embedded feature distribution alignment |
| CN111259720A (en) * | 2019-10-30 | 2020-06-09 | 北京中科研究院 | Unsupervised pedestrian re-identification method based on self-supervision agent feature learning |
| CN111931619A (en) * | 2020-07-31 | 2020-11-13 | 杭州电子科技大学 | Cross-domain gait re-identification method based on optimized clustering algorithm |
| CN112183581A (en) * | 2020-09-07 | 2021-01-05 | 华南理工大学 | Semi-supervised mechanical fault diagnosis method based on self-adaptive migration neural network |
| CN112487190A (en) * | 2020-12-13 | 2021-03-12 | 天津大学 | Method for extracting relationships between entities from text based on self-supervision and clustering technology |
| CN112699892A (en) * | 2021-01-08 | 2021-04-23 | 北京工业大学 | Unsupervised field self-adaptive semantic segmentation method |
| CN112729835A (en) * | 2021-01-30 | 2021-04-30 | 温州大学 | Multi-domain semi-supervised fault diagnosis method and device for axial plunger pump bearing |
| CN113191183A (en) * | 2021-03-02 | 2021-07-30 | 清华大学 | Unsupervised domain false label correction method and unsupervised domain false label correction device in personnel re-identification |
| CN113240506A (en) * | 2021-05-10 | 2021-08-10 | 浙江大学 | Financial wind-controlled cold start modeling method based on unsupervised field self-adaptation |
| CN113255573A (en) * | 2021-06-17 | 2021-08-13 | 成都东方天呈智能科技有限公司 | Pedestrian re-identification method based on mixed cluster center label learning and storage medium |
| CN113255823A (en) * | 2021-06-15 | 2021-08-13 | 中国人民解放军国防科技大学 | Unsupervised domain adaptation method and unsupervised domain adaptation device |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11625557B2 (en) * | 2018-10-29 | 2023-04-11 | Hrl Laboratories, Llc | Process to learn new image classes without labels |
| US11366988B2 (en) * | 2019-05-28 | 2022-06-21 | Wipro Limited | Method and system for dynamically annotating and validating annotated data |
| CN110555273B (en) * | 2019-09-05 | 2023-03-24 | 苏州大学 | Bearing life prediction method based on hidden Markov model and transfer learning |
-
2021
- 2021-08-18 CN CN202110949934.XA patent/CN113792758B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102039138B1 (en) * | 2019-04-02 | 2019-10-31 | 주식회사 루닛 | Method for domain adaptation based on adversarial learning and apparatus thereof |
| CN110705525A (en) * | 2019-10-24 | 2020-01-17 | 北京建筑大学 | Method and device for diagnosing rolling bearing fault |
| CN111259720A (en) * | 2019-10-30 | 2020-06-09 | 北京中科研究院 | Unsupervised pedestrian re-identification method based on self-supervision agent feature learning |
| CN111144458A (en) * | 2019-12-13 | 2020-05-12 | 重庆交通大学 | Method for identifying mechanical faults under different working conditions of subspace embedded feature distribution alignment |
| CN111144565A (en) * | 2019-12-27 | 2020-05-12 | 中国人民解放军军事科学院国防科技创新研究院 | Self-supervision field self-adaptive deep learning method based on consistency training |
| CN111931619A (en) * | 2020-07-31 | 2020-11-13 | 杭州电子科技大学 | Cross-domain gait re-identification method based on optimized clustering algorithm |
| CN112183581A (en) * | 2020-09-07 | 2021-01-05 | 华南理工大学 | Semi-supervised mechanical fault diagnosis method based on self-adaptive migration neural network |
| CN112487190A (en) * | 2020-12-13 | 2021-03-12 | 天津大学 | Method for extracting relationships between entities from text based on self-supervision and clustering technology |
| CN112699892A (en) * | 2021-01-08 | 2021-04-23 | 北京工业大学 | Unsupervised field self-adaptive semantic segmentation method |
| CN112729835A (en) * | 2021-01-30 | 2021-04-30 | 温州大学 | Multi-domain semi-supervised fault diagnosis method and device for axial plunger pump bearing |
| CN113191183A (en) * | 2021-03-02 | 2021-07-30 | 清华大学 | Unsupervised domain false label correction method and unsupervised domain false label correction device in personnel re-identification |
| CN113240506A (en) * | 2021-05-10 | 2021-08-10 | 浙江大学 | Financial wind-controlled cold start modeling method based on unsupervised field self-adaptation |
| CN113255823A (en) * | 2021-06-15 | 2021-08-13 | 中国人民解放军国防科技大学 | Unsupervised domain adaptation method and unsupervised domain adaptation device |
| CN113255573A (en) * | 2021-06-17 | 2021-08-13 | 成都东方天呈智能科技有限公司 | Pedestrian re-identification method based on mixed cluster center label learning and storage medium |
Non-Patent Citations (6)
| Title |
|---|
| "A new method for intelligent fault diagnosis of machines based on unsupervised domain adaptation";Lu N等;《Neurocomputing》;第427卷;第96-109页 * |
| "Domain Adaptation by Class Centroid Matching and Local Manifold Self-Learning";Tian L等;《arxiv》;第1-12页 * |
| "Joint unsupervised learning of deep representatins and image clusters";Yang J等;《Proceedings of the IEEE conference on computer vision and pattern recognition》;第5147-5156页 * |
| "基于局部语义特征不变性的跨域行人重识别";张晓伟等;《北京航空航天大学学报》;第46卷(第9期);第1682-1690页 * |
| "基于滑动窗口密度聚类的数据流偏倚采样算法";胡志冬等;《计算机科学》;第40卷(第9期);第254-256页及第269页 * |
| "模糊聚类在深度学习自监督领域的应用";范少帅等;《中国新通信》;第23卷(第4期);第166-167页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113792758A (en) | 2021-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113792758B (en) | Rolling bearing fault diagnosis method based on self-supervision learning and clustering | |
| CN111651937B (en) | Method for diagnosing faults of in-class self-adaptive bearing under variable working conditions | |
| CN110555273B (en) | Bearing life prediction method based on hidden Markov model and transfer learning | |
| CN110361176B (en) | Intelligent fault diagnosis method based on multitask feature sharing neural network | |
| CN112629863B (en) | Bearing fault diagnosis method for dynamic joint distribution alignment network under variable working conditions | |
| CN109580215B (en) | Wind power transmission system fault diagnosis method based on deep generation countermeasure network | |
| CN113935406B (en) | Mechanical equipment unsupervised fault diagnosis method based on countermeasure flow model | |
| CN114048568B (en) | Rotary machine fault diagnosis method based on multisource migration fusion shrinkage framework | |
| CN112763214B (en) | Rolling bearing fault diagnosis method based on multi-label zero-sample learning | |
| CN110647830A (en) | Bearing fault diagnosis method based on convolutional neural network and Gaussian mixture model | |
| CN114004252A (en) | Bearing fault diagnosis method, device and equipment | |
| CN113076834B (en) | Rotating machine fault information processing method, processing system, processing terminal, and medium | |
| CN114417913B (en) | Bearing life prediction method based on PC-TCN and migration learning | |
| CN113008559A (en) | Bearing fault diagnosis method and system based on sparse self-encoder and Softmax | |
| CN114118138A (en) | Bearing composite fault diagnosis method based on multi-label field self-adaptive model | |
| CN116793682A (en) | Bearing fault diagnosis method based on iCORAL-MMD and anti-migration learning | |
| CN114491828A (en) | Compound fault diagnosis method based on multi-gating hybrid expert multi-task learning | |
| CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
| CN116186641A (en) | Cross-equipment rolling bearing small sample fault diagnosis method based on feature fusion | |
| CN114564987A (en) | Rotary machine fault diagnosis method and system based on graph data | |
| CN113076920B (en) | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model | |
| Yang et al. | A Balanced Deep Transfer Network for Bearing Fault Diagnosis | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| CN115876467A (en) | Pseudo label transfer type two-stage field self-adaptive rolling bearing fault diagnosis method | |
| CN115496941A (en) | Knowledge-enhanced computer vision-based structural health diagnosis method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |